Threat Modeling AI Agents (Prompt → Tool → Prod)

Most AI incidents aren’t “model hacks.” They’re workflow hacks — where prompts, tools, permissions, and data flows collide.

You’ll learn:

• The “Prompt → Tool → Action” chain: how it breaks in production
• Attacker goals: data exfil, privilege abuse, transaction fraud, lateral movement
• Threat modeling patterns for RAG, connectors, agents, IDE copilots, and MCP/tool plugins
• Control mapping: what to log, restrict, test, and continuously validate

Agenda:

• Welcome + context
• Deep dive talk
• Workshop: build a threat model for a real AI assistant (hands-on)
• Q&A + networking

Join the community (free):

• Slack (free workshops + closed community calls + Open Source): https://join.slack.com/t/aisecuritycommunity/shared_invite/zt-3l88a89lw-NvdP6d9Wa0zGLxsv8aSk7Q
• WhatsApp (announcements + quick updates + Opportunities): https://chat.whatsapp.com/CQoDbFi4V8jAxgYgBBQPk7

Call for Proposals (CFP) — Meetups + June Conference:
We’re opening CFP for speakers/workshops/panels for both:

• Monthly meetups (Pune & Bangalore)
• AI Security Conference (June 2026)

Submit your talk/workshop idea on #cfp channel in slack.

Suggested themes: agent security, IDE/copilot security, MCP/tooling security, red teaming, governance & compliance, detection/IR, secure AI SDLC, case studies.

In association with:
1. Guard0: https://guard0.ai
2. More coming soon.